Skip to main content
2024

Nubank adopts AWS Graviton processing and optimizes costs by 14%

An AWS customer since 2013, Nubank is migrating all of its critical workloads to the AWS Graviton processor, in a process that, so far, has resulted in a 14% reduction in cloud costs.

Benefits

14%

reduction in cloud usage costs

21%

reduction in carbon emissions

Construction

of a more flexible architecture

Overview

Known as one of the largest digital financial service platforms in the world, Nubank now has around 94 million customers combined in total across its Brazil, Mexico and Colombia operations. Conceived as a digital-first institution, the institution was established in 2013, operating entirely on the cloud from day one. “Our structure has remained with AWS from day one. It's a long-standing relationship,” explains Nubank's engineering director, Vitor Monteiro Puente.

Nubank’s DNA is more than using the cloud. It is also a constant quest for cost savings and performance gains seeking to streamline operations and make them increasingly simpler for its customers. “This applies to all areas of the institution,” says Puente, recalling that this has led to a series of initiatives, both internal and together with AWS, always seeking to optimize cloud costs. “The migration to AWS Graviton processors was born out of this quest,” he says.

Missing alt text value

About Nubank

Created in 2013, Nubank is now one of the largest digital financial services platforms in the world, serving more than 90 million customers across Brazil, Mexico, and Colombia. The bank uses technology to create new financial solutions and experiences.

Opportunity | In Search of more Efficiency and Lower Costs

Created by AWS to provide the best price-performance ratio for workloads running on Amazon Elastic Compute Cloud (Amazon EC2), AWS Graviton processors began to be assessed by Nubank in 2021. Since then, the institution has been following the evolution of these processors, as well as AWS services compatibility for Amazon ElastiCache, Amazon RDS and OpenSearch, for example. According to Puente, this is when the first experiments with the processors began.

One of his team's premises, according to him, is to define the simplest and easiest ways to test hypotheses. In the case of AWS Graviton processors, testing began with some microservices and databases. “That’s how we started. Then we expanded the scope. In some cases, we were forced to revisit our infrastructure after testing—a process of experimentation and learning that allowed us to achieve full impact across different applications,” he says.

Nubank's systems engineer, Carlos Diogo, recalls that the first phase of testing included a benchmark to understand how the new processors would work. “They were processing twice the volume with the same latency,” he reveals.

The testing phase lasted about a year, after which the Nubank IT team decided to enable its entire structure for the new processors. Once the decision was made, a new scaled rollout was planned, which actually began in the second half of 2022. “Our first machines started migrating in October of that year in a process that is still ongoing,” he says, mentioning the data environment, which was migrated in three weeks, as an example.

Why AWS | Planned and Performed Migration

Nubank's product operations lead, Daniel Capp, explains that the migration revealed some challenges, such as prioritizing the adoption of the new processors with the Nubank teams and rebuilding some plugins. He recalls that the Data Analytics team, for example, underwent internal training before the migration. “Today we ran the entire Analytics workload on AWS Graviton,” he says.

Diogo stresses that the migration process brought some very positive surprises, such as the absence of runtime problems. “We also had no problems with the service libraries or with the Amazon ElastiCache migration,” he says.

According to Puente, today the entire Nubank microservices environment is enabled to work with AWS Graviton. “In our storage layer, part of the workload already uses Graviton, but the Datomic database is still being validated. There are other very specific workloads,” he explains, remembering that there will always be one that connects in a specific way, or to a specific software. For these demands, Nubank and AWS are working on a case-by-case basis.

“We are taking it slowly with specific services. For example, our Pix feature is more real time and uses a very specific architecture. We have service level agreements (SLAs) with the Central Bank that cannot be broken. These nuances have to be worked on cautiously”, she reveals, remembering that caution is important to maintaining service stability. This is the phase in which the migration process is today: defining what Puente calls specific points and then migrating the latest services. In any case, today the institution has more than 90% of its services enabled for or running on AWS Graviton.

Results | Successfully Optimized Costs

Despite it is still migrating, Nubank is already able to see results from using its services with AWS Graviton processors. Puente reveals that, so far, there has been a 14% optimization in cloud usage costs.

On another front, Nubank further reduced its carbon footprint. The institution, which has had negative emissions since 2020, reduced emissions by another 21% by using the new processors. In addition, the institution was able to develop a flexible architecture that allows it to use whichever processors are most efficient based on demand. “Our architecture today runs on both AWS Graviton and x86,” Puente reveals. The use of this multi-architecture model allows the institution to achieve benchmarks and run its workload on the one that is most efficient, taking into account cost and stability.

With these results, Nubank is already preparing for the use of third generation processors, which will bring improvements compared to the results obtained so far. “The changes we made allow us to use the AWS Graviton versions and that’s why we are looking forward to AWS Graviton 3, which should increase our efficiency,” he says, revealing that tests of the new version are already underway with promising results in the storage layer, for example.

Purple Netflix logo with stylized text.
There was an increase in the use of Graviton in the Data Analytics environment from 10% to 57%, reducing the carbon footprint by 21%, according to an AWS study.

Vitor Monteiro Puente

Director of Engineering at Nubank

AWS Services Used

AWS Graviton

AWS Graviton is a family of processors designed to provide the best cost-benefit for your cloud workloads running on Amazon Elastic Compute Cloud (Amazon EC2). Choose the instance based on which AWS Graviton best meets your needs.

Learn more

Amazon ElastiCache

Amazon ElastiCache is a serverless cache service, compatible with Redis and Memcache, that offers cost-effective, real-time performance for modern applications. 

Learn more

Amazon EC2

Amazon Elastic Compute Cloud (Amazon EC2) offers the broadest and most in-depth computing platform, with more than 750 instances and options for the latest processors, storage, networks, operational systems, and purchasing models to help you better address the needs of your workload.

Learn more

Amazon RDS

Amazon RDS is an easy-to-manage relational database service optimized for total cost of ownership.

Learn more

Get Started

Organizations of all sizes across all industries are transforming their businesses and delivering on their missions every day using AWS. Contact our experts and start your own AWS journey today.

Contact Sales